Key Takeaways
- Northwestern University researchers found TikTok's algorithm responds to user feedback but only temporarily.
- Negative signals like 'not interested' do not consistently remove disliked content from the For You Page.
- The study highlights the limitations of user control over personalized content feeds.
A recent study by Northwestern University computer scientists has revealed that TikTok's algorithm, which powers its For You Page (FYP), does not always respect users' negative feedback as effectively as advertised. The research, published in Ars Technica, challenges the notion of user agency over content selection on the platform.
The FYP is a personalized content feed driven by an algorithm that uses both implicit and explicit signals to determine what videos users see next. Implicit signals include how long users watch specific videos, while explicit signals encompass likes, follows, and direct feedback such as 'not interested' prompts.
According to the study, when users indicate they are not interested in a particular video or type of content, this negative signal does have an impact on future recommendations. However, the effect is temporary. The algorithm gradually reverts to its default behavior unless consistent negative feedback is provided over time.
Piotr Sapiezynski, one of the co-authors of the study and a specialist in 'algorithm audits,' explained that these findings are significant because they highlight the limitations of user control on online platforms. 'How they work, how they fail, when they fail, how they harm individuals and societies'—these are the questions his research group aims to answer.
The implications of this study extend beyond just TikTok. It raises broader concerns about the effectiveness of feedback mechanisms in algorithm-driven content feeds across various social media platforms. Users may feel empowered by these features but might not have as much control over their experience as they believe.
Sapiezynski noted, 'On the other hand, it's unclear why the platforms would offer it if it doesn't work.' This statement underscores the potential mismatch between user expectations and platform capabilities. The research suggests that while negative feedback can influence content recommendations temporarily, consistent and persistent user actions are necessary to see long-term changes.
The findings of this study have important implications for users who rely on these features to curate their online experiences. It may prompt a reevaluation of how much control individuals actually have over the content they encounter on social media platforms like TikTok.
'On the other hand, it's unclear why the platforms would offer it if it doesn't work.'
Piotr Sapiezynski, Co-author of the study and specialist in 'algorithm audits'
